Apppublic

Mohammed2311/Azure_Cost_Estimation_Agent

sourceHugging Facemitupdated 10mo agoView on Hugging Face

0likes

App README

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference -----

☁️ Azure Cost Estimation Agent

Submission for the MCP 1st Birthday Hackathon

A production-ready agentic system that transforms natural language architecture descriptions into accurate, real-time Azure cost estimates. By leveraging the Model Context Protocol (MCP), this agent connects a Gemini reasoning engine directly to a local Azure Pricing tool server, replacing manual spreadsheet work with intelligent conversation.

💡 The Problem & Solution

The Problem: Turning an architectural idea (e.g., "I need a mobile backend for 10k users") into a cost estimate is tedious. It requires hunting through pricing pages, finding specific SKUs, and manually calculating monthly rates.

The Solution: An Agentic Workflow:

1.Reasoning: The agent interprets your description and infers necessary resources (App Service, Cosmos DB, Bandwidth, etc.).
2.Tool Use (MCP): It autonomously calls a local MCP server (azure_pricing_mcp_server.py) to fetch real-time retail prices and discover SKUs.
3.Visualization: It aggregates the data into a clean report with interactive Plotly charts.

✨ Features

—Natural Language Input: Just describe your infrastructure requirements.
—Real-Time Pricing: No hardcoded values. The agent queries the Azure Retail Prices API via MCP.
—ReAct Pattern: Uses a "Thought → Action → Observation" loop to refine its search and calculations.
—Interactive UI: Built with Gradio for a chat-like experience.
—Visualizations: Generates Pie and Bar charts using Plotly to visualize cost distribution.
—Transparent Logic: Displays the full "Agent Analysis Trace" so you can see exactly what tools were called.

🏗️ Architecture

The system consists of two main components communicating via the Model Context Protocol over stdio:

1.Client (`session.py`): The Gradio frontend and Gemini Agent. It sends prompts and tool requests.
2.Server (`azure_pricing_mcp_server.py`): A standalone script that implements MCP tools (azure_price_search, azure_cost_estimate) to fetch data from Azure.

mermaid

graph LR
    User[User] -->|Input| UI[Gradio Interface]
    UI -->|Prompt| Gemini[Gemini 2.5 Flash Lite]
    
    subgraph "MCP Connection (Stdio)"
        Gemini <-->|ReAct Loop| AgentLogic
        AgentLogic <-->|JSON-RPC| MCPServer[Azure Pricing MCP Server]
    end
    
    MCPServer <-->|HTTP GET| AzureAPI[Azure Retail Prices API]

🛠️ Installation

Prerequisites

—Python 3.10 or higher
—A Google Gemini API Key (Get one here)

1\. Clone the Repository

bash

git clone https://github.com/your-username/azure-cost-agent.git
cd azure-cost-agent

2\. Set up Virtual Environment

bash

python -m venv venv
# Windows
.\venv\Scripts\activate
# Mac/Linux
source venv/bin/activate

3\. Install Dependencies

Create a requirements.txt file (or run the command below):

bash

pip install mcp google-generativeai gradio plotly python-dotenv requests

4\. Configure Environment

Create a .env file in the root directory and add your key:

env

GEMINI_API_KEY=your_actual_gemini_api_key_here

5\. Verify Files

Ensure both Python scripts are in the same folder:

—session.py (The main application)
—azure_pricing_mcp_server.py (The MCP server script)

🚀 Usage

Run the main session file. The MCP server will be started automatically as a subprocess.

bash

python session.py

Once running, click the local URL provided (usually http://127.0.0.1:7860).

Example Prompts to Try

—"I need a mobile app backend with REST APIs, authentication, and MongoDB. Expected 10,000 active users in Europe."
—"Deploy a data analytics platform: batch processing 8h/day, 2TB data warehouse, ML training twice weekly."
—"Simple WordPress site on a VM with a MySQL database in East US."

🧩 Technical Details

The Tools

The agent has access to the following MCP tools:

—azure_discover_skus: Finds available SKUs for a specific service and region.
—azure_price_search: Searches for specific pricing meters.
—azure_cost_estimate: Calculates monthly costs based on usage patterns (hours/day, quantity).
—azure_price_compare: Compares costs across different Azure regions.

Technologies Used

—Model Context Protocol (MCP): For standardized tool definitions and client-server communication.
—Google Gemini 2.5 Flash Lite: For fast, accurate reasoning and JSON parsing.
—Gradio: For the web interface.
—Plotly: For dynamic data visualization.

🤝 Contributing

Contributions are welcome\! Please fork the repository and submit a pull request.

📄 License

Distributed under the MIT License. See LICENSE for more information.